Skip to content
Computer vision

CCA Traj Qwen3.8-27B v12

27B-модель с 128k контекстом для агентных траекторий

1 min read

Это full-parameter SFT от rqadri на базе Qwen/Qwen3.8-27B: 27,36 млрд параметров, контекст 128k, обучение на траекториях кодирующих агентов CCA. Артефакт исследовательский — модель ещё не оценена.

Ключевые особенности

  • Full-parameter обучение на CCA coding-agent trajectories: v1 + v2, всего 1 982 последовательности.
  • 128k контекст, bf16, DeepSpeed ZeRO-3 с CPU offload и Ulysses sequence parallelism на Intel XPU (Aurora).
  • Loss считается только на span'ах ассистента; один обучающий пример — целая генерация контекста агента.
  • Данные: cca-large-v1 (1 243 последовательности, 61,9M токенов) и cca-large-v2 (739 последовательностей, 47,7M токенов), только open-weight teachers.
  • Для инференса нужно грузить как Qwen3_5ForConditionalGeneration; в комплекте есть preprocessor_config.json и video_preprocessor_config.json для vLLM.

Пример

Сценарий из карточки: агенту подаётся длинная траектория кодирующего агента — до 128k токенов, — а модель учится предсказывать только ответы ассистента. Для запуска используйте класс Qwen3_5ForConditionalGeneration: текстовая версия инициализирует случайные веса.

ПодробнееПрогон: lr 1e-5, 1 эпоха = 248 шагов; остановлен на шаге 244/248, lr 1,07e-8, последний loss 0,447. Веса — consolidated checkpoint-244 в bf16.
Пока не оценена: бенчмарков в карточке нет. Это исследовательский артефакт, а не готовый продакшн-чекпоинт.
Железо обучения: Intel XPU / Aurora, DeepSpeed ZeRO-3, 8 ranks Ulysses.
Discuss Telegram 0 18

Will it run?

Working out the memory needs

The bot is still collecting this model's weight sizes; the estimate will appear here in a few minutes. Meanwhile you can rent a GPU with headroom.

Massed ComputeOpen Massed Compute

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # rqadri/cca-traj-qwen38-27b-v12
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download rqadri/cca-traj-qwen38-27b-v12 --local-dir ~/models/cca-traj-qwen38-27b-v12
    # The model is in ~/models. How to run it is in the model card (link above).
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of cca-traj-qwen38-27b-v12 (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

rqadri/cca-traj-qwen38-27b-v12
Author
rqadri
Category
Computer vision
Released
Posted

On NeyroLearn

Hugging Face

180

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments